Wetland Health and Carbon Stocks Monitoring Using Advanced Machine Learning Algorithms and Remote Sensing Data

点击翻译按钮获取中文摘要

Over the past two decades, the value of Earth Observation (EO) data has been well documented for many environmental applications, such as forest and wetland monitoring, using conventional remote sensing techniques. Although these techniques are of great useful, data availability, great dependency on feature engineering and manual editing, as well as insufficient accuracy, limited their applications to small-scale study areas. However, most recently, advances in remote sensing technologies, artificial intelligence (AI) algorithms, and powerful cloud computing infrastructure, such as Google Earth Engine (GEE), have addressed the limitations of traditional techniques. Despite the recent advances, exploiting these new methodologies for wetland monitoring has remained limited due to the lack of sufficient research and development with respect to the following aspects: removing or reducing barriers related to the need for large volume of ground truth data; advancing toward new technologies that decrease the level of human involvement in data processing; enhancing techniques of data processing to raise the level of accuracy and interpretability of the outputs; and moving toward geo big data analysis for large-scale investigations. In this regard, the long-term objective of the applicant's research program (10+ years) is to develop a state-of-the-art open-access wetland monitoring toolbox at a national scale using advanced AI tools and remote sensing data to support Canada's rich biodiversity. To meet the long-term research goal, this short-term research program will address the challenges related to the development of the wetland monitoring toolbox. Particularly, the thematic application of the proposed solutions falls within three major important categories, namely wetland classification, wetland change detection, and wetland above ground biomass (AGB) estimation, all of which are essential for wetland protection and management. The outcomes of the proposed research have the potential to significantly advance and revolutionize the techniques of wetland mapping and monitoring, particularly at a large scale. The wetland monitoring toolbox developed in this research program will capture comprehensive knowledge about the status of wetlands at a provincial scale with a great capacity to upgrade at the national scale and beyond the borders of Canada, thus contributing to global-scale wetland knowledge and the improvement of both national and global agreements and policies. Finally, the achievements of this program can also be easily transferred to other domains of environmental monitoring, such as forest monitoring. The highly qualified personnel (HQP) will be trained in a multidisciplinary environment and will be able to transfer their knowledge and skills to Canadian academia and industries in the wide variety of applications for mapping and monitoring Canada's natural resources using state-of-the-art geospatial technologies.